Mahmoud SM, Mohamed MM, Abdelfatah RZE, Abdelkreem E. Comparison of six neonatal mortality prediction scores: A prospective observational study. World J Clin Pediatr 2026; 15(4): 117203 [DOI: 10.5409/wjcp.117203]
Corresponding Author of This Article
Elsayed Abdelkreem, MD, PhD, Associate Professor, Consultant, DHPE, Department of Pediatrics, Faculty of Medicine, Sohag University, Sohag University Street, Nasser City 82524, Sohag, Egypt. d.elsayedmohammed@med.sohag.edu.eg
Research Domain of This Article
Pediatrics
Article-Type of This Article
research-article
Open-Access Policy of This Article
This article is an open-access article which was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Baishideng Publishing Group Inc, 7041 Koll Center Parkway, Suite 160, Pleasanton, CA 94566, USA
Share the Article
Mahmoud SM, Mohamed MM, Abdelfatah RZE, Abdelkreem E. Comparison of six neonatal mortality prediction scores: A prospective observational study. World J Clin Pediatr 2026; 15(4): 117203 [DOI: 10.5409/wjcp.117203]
Shaimaa Mohamed Mahmoud, Montaser Mohamed Mohamed, Rehab Zein Elabdeen Abdelfatah, Elsayed Abdelkreem, Department of Pediatrics, Faculty of Medicine, Sohag University, Nasser City 82524, Sohag, Egypt
Elsayed Abdelkreem, Department of Clinical Sciences, Al-Rayan National College of Medicine, Al-Rayan National Colleges, Al Madinah Al Munawwarah 41411, Saudi Arabia
Author contributions: Mahmoud SM, Mohamed MM, and Abdelfatah RZE contributed to edit the manuscript; Mahmoud SM, Mohamed MM, Abdelfatah RZE, and Abdelkreem E contributed to methodology; Mahmoud SM, Abdelfatah RZE, and Abdelkreem E contributed to data analysis and interpretation; Mohamed MM and Abdelkreem E shared in study design; Abdelfatah RZE contributed to investigation; Abdelkreem E wrote the manuscript draft. All authors have read and approved the final manuscript.
Institutional review board statement: This study was approved by the Research Ethics Committee of Faculty of Medicine, Sohag University (approval No. Soh-Med-23-03-10MS; dated March 08, 2023).
Informed consent statement: All study participants, or their legal guardian, provided informed written consent prior to study enrollment.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
STROBE statement: The authors have read the STROBE Statement-checklist of items, and the manuscript was prepared and revised according to the STROBE Statement-checklist of items.
Data sharing statement: The data that support the findings of this study are available from the corresponding author upon reasonable request.
Corresponding author: Elsayed Abdelkreem, MD, PhD, Associate Professor, Consultant, DHPE, Department of Pediatrics, Faculty of Medicine, Sohag University, Sohag University Street, Nasser City 82524, Sohag, Egypt. d.elsayedmohammed@med.sohag.edu.eg
Received: December 1, 2025 Revised: December 27, 2025 Accepted: March 13, 2026 Published online: December 9, 2026 Processing time: 312 Days and 8.1 Hours
Abstract
BACKGROUND
Various scoring systems have been developed for early identification of seriously ill neonates in the neonatal intensive care unit, but studies comparing their predictive accuracies are scarce.
AIM
To compare the performance of six neonatal mortality prediction scores in a resource-limited setting.
METHODS
This prospective observational study included neonates admitted within 24 hours of birth to a major Egyptian neonatal intensive care unit from March 2023 to May 2024. Each newborn was assessed within 12 hours after admission for Clinical Risk Index for Babies II (CRIB II), Score for Neonatal Acute Physiology II, Score for Neonatal Acute Physiology Perinatal Extension II, sensorium, temperature, oxygenation, perfusion, skin color, and blood sugar (STOPS), Modified Sick Neonatal Score (MSNS), and neonatal sequential organ failure assessment (nSOFA). Area under the receiver operating characteristic curve (AUC) was calculated for each score to assess and compare their predictive accuracy for all-cause in-hospital mortality.
RESULTS
A total of 299 neonates were included. Compared with survivors (n = 194), nonsurvivors (n = 105) exhibited significantly lower MSNS and higher other scores. The AUCs values were MSNS (0.91) > Score for Neonatal Acute Physiology Perinatal Extension II and STOPS (each 0.89) > neonatal sequential organ failure assessment and Score for Neonatal Acute Physiology II (each 0.88) > CRIB II (0.82). Overall predictive accuracy ranged from 87.3% for both MSNS (cutoff < 10) and STOPS (cutoff > 2) to 75.3% for CRIB II (cutoff > 4). All AUCs were comparable, except for CRIB II, which performed better in preterm babies < 33 weeks than in the overall cohort.
CONCLUSION
This study demonstrated good and comparable performance of multiple neonatal mortality prediction scores in a resource-limited setting, with more simple scores such as MSNS and STOPS offering the advantages of greater applicability.
Core Tip: This study provides the first head-to-head comparison of six neonatal mortality prediction scores applied within the same neonatal intensive care unit population in a resource-limited, high-mortality setting. All models demonstrated good-to-excellent discriminatory performance. Importantly, simple bedside scores, such as the Modified Sick Neonatal Score and sensorium, temperature, oxygenation, perfusion, skin color, and blood sugar score performed comparably to more complex physiology-based tools. These findings highlight the value of feasible, rapid, and non-invasive scoring systems for early risk stratification and clinical decision-making in neonatal intensive care units in low- and middle-income settings.
Citation: Mahmoud SM, Mohamed MM, Abdelfatah RZE, Abdelkreem E. Comparison of six neonatal mortality prediction scores: A prospective observational study. World J Clin Pediatr 2026; 15(4): 117203
In 2023, approximately 2.3 million newborns died globally, with the majority occurring in low- and middle-income countries (LMICs). The leading causes remain preterm birth, perinatal asphyxia/trauma, neonatal infections, and congenital anomalies[1-4]. Various scoring systems have been developed for early identification of seriously ill neonates in the neonatal intensive care unit (NICU)[5,6]. These scores are critical for prioritization of care and effective resource allocation, particularly in resource-limited settings, which improves clinical outcomes[7,8]. Moreover, they serve as tools for risk adjustment and quality assessment, enabling benchmarking of NICU performance across time and institutions, and guiding healthcare policy decisions[7,9,10]. These tools typically contain a combination of demographic, clinical, and laboratory variables, such as gender, gestational age, birth weight, temperature, conscious level, perfusion, oxygenation status, and acid-base balance[8,11]. Notable examples include Clinical Risk Index for Babies II (CRIB II)[12], Score for Neonatal Acute Physiology II (SNAP-II) and its SNAP Perinatal Extension II (SNAPPE-II)[13], Sensorium, temperature, oxygenation, perfusion, skin color, and blood sugar (STOPS)[11], Modified Sick Neonatal Score (MSNS)[9], and neonatal sequential organ failure assessment (nSOFA)[10].
The predictive accuracy of individual scores for neonatal mortality has been investigated in several studies, but comparative studies are scarce[5,14]. Gagliardi et al[15] and Vardhelli et al[16] showed that CRIB II outperforms SNAPPE-II in very low birthweight (VLBW) neonates. A subsequent study by Vardhelli et al[11] reported better performance for SNAPPE-II over STOPS in newborns ≥ 33 gestational weeks. Hao et al[17] described that nSOFA had superior predictive accuracy compared to SNAPPE-II and CRIB-II in preterm babies. A systematic review by Veloso et al[5] identified SNAPPE-II as the most accurate tool in the general NICU population, while another systematic review by Zeng et al[14] highlighted CRIB and SNAPPE-II as top performers among preterm infants. However, neither review evaluated newly emerged tools such as STOPS, MSNS, and nSOFA, which offer greater feasibility in resource-limited settings[11]. Importantly, no study has yet conducted a direct head-to-head comparison of STOPS, MSNS, and nSOFA with each other and with traditional scores such as SNAPPE-II and CRIB II within a single population.
Given these knowledge gaps, the authors performed a prospective observational study to compare the accuracy of six neonatal mortality prediction scores - CRIB II, SNAP-II, SNAPPE-II, STOPS, MSNS, and nSOFA - in a single cohort of NICU admissions in a resource-limited setting. The study provides novel comparative insights and helps inform the selection of appropriate scoring systems in LMICs.
MATERIALS AND METHODS
Study design and setting
This prospective observational study was conducted from March 2023 to May 2024 at Sohag University Hospital, a major referral center in southern Egypt that serves a population of six million. The NICU admits an average of 500 newborns annually and is equipped with 36 incubators, 20 hemodynamic monitors, and 12 ventilators. High-frequency oscillatory ventilation and extracorporeal membrane oxygenation are not currently available, and the use of surfactant therapy is primarily limited to preterm babies with respiratory distress syndrome. Laboratory tests such as arterial blood gas analysis are not consistently available; however, special arrangements were made to ensure these essential tests were performed for all newborns throughout the study period. This study was approved by the Research Ethics Committee of Faculty of Medicine, Sohag University (approval No. Soh-Med-23-03-10MS; dated March 08, 2023). Written informed consent was obtained from parents or legally authorized representatives of all participating neonates. All study procedures were performed in compliance with the ethical principles enclosed in the 1964 Declaration of Helsinki and its later revisions. The manuscript adhered to Strengthening the Reporting of Observational Studies in Epidemiology guidelines for reporting prospective observational studies.
Participants
Neonates admitted within the first 24 hours of life, irrespective of gestational age and birth weight, were eligible for inclusion. Exclusion criteria were failure to obtain informed consent, major life-threatening congenital anomalies (e.g., tracheoesophageal fistula, complex congenital heart diseases, neural tube defects, and omphalocele), death within 12 hours after admission, transfer to another hospital, discharge against medical advice, and missing data for scoring systems.
Sample size calculation
We calculated the sample size using MedCalc for Windows, version 19.4 (MedCalc Software Ltd, Ostend, Belgium) to detect a 10% difference in the area under the receiver operating characteristic curve (AUC) among scores for predicting in-hospital neonatal mortality. Assuming AUC values of 0.80 and 0.90, a survival-to-mortality ratio of 3:1, type 1 error of 0.05, and 90% power, the minimum required sample size was 276 newborns.
Assessment and outcomes
For each enrolled neonate, six scores (CRIB II, SNAP-II, SNAPPE-II, STOPS, MSNS, and nSOFA) were calculated simultaneously by the same physicians using the worst recorded parameters in the first 12 hours after NICU admission. The physicians were trained before initiation of the study and were supervised by the investigators to maintain precision and uniformity in the calculations. Details of each scoring system are provided in Supplementary material 1. In brief, CRIB II (range 0 to 27) includes five parameters: Sex, gestational age, birth weight, body temperature at admission, and worst base deficit in the first hour of life[12]. SNAP-II (range 0 to 115) contains six postnatal variables: Mean blood pressure, lowest body temperature, pO2/FiO2 ratio, lowest serum PH, multiple seizures, and urine output[13]. SNAPPE-II (range 0 to 162) adds three perinatal factors to SNAP-II, namely birth weight, small for gestational age, and APGAR score at 5 minutes[13]. STOPS (range 0 to 10) is calculated based on six components: Sensorium, temperature, oxygenation, perfusion, skin color, and random blood sugar[11]. MSNS (range 0 to 16) relies on eight items: Respiratory effort, heart rate, axillary temperature, capillary refill time, random blood sugar, SpO2, gestational age, and birth weight[9]. Finally, nSOFA (range 0 to 15) incorporates three subscores for respiratory (SpO2:FiO2 ratio and presence of mechanical ventilation), cardiovascular (use of vasoactive agents and systemic corticosteroids), and hematologic (presence and degree of thrombocytopenia) dysfunctions[10].
Gestational age was determined by the last menstrual period along with first-trimester obstetric ultrasound and was verified using the new Ballard score. Weight was measured by a digital weighing scale (Rossmax International Ltd, Taipei, Taiwan, China) with precision to the nearest gram. Small for gestational age was defined following 2013-revised Fenton growth charts[18]. Vital signs and SpO2 were measured using Vista 120 hemodynamic monitor (Dräger Medical Inc., Lübeck, Germany). Capillary refill time was recorded by a stopwatch as the time for return of the skin color over sternum after removal of 5-second pressure by index finger. Urine output was estimated by weighing diapers pre- and post-urination. Random blood glucose was checked by ACCU-CHEK glucometer (Roche Diabetes Care Ltd, Mumbai, India) through test strips. Blood counts were assessed by Sysmex XN-1000TM Hematology Analyzer (Sysmex Corp., Kobe, Japan). Serum pH, base excess, and pO2 were measured using ABL800 FLEX blood gas analyzer (Radiometer Medical ApS, Bronshoj, Denmark). For neonates on respiratory support, FiO2 was recorded at the same time of arterial blood gas analysis. All neonates received the standard of care according to institutional protocol and national guidelines and were monitored until discharge or death. The primary outcome of this study was all-cause in-hospital mortality.
Statistical analysis
Data were analyzed using Stata/BE 17 (StataCorp LLC, College Station, TX, United States) and were presented as n (%), mean ± SD, and median (interquartile range), as appropriate. We compared variables between non-survivors and survivors using χ2/Fisher exact test for categorical data, independent t-test for normally distributed quantitative measures, and Wilcoxon rank-sum test for non-normally distributed quantitative data. Receiver operating characteristics curve analysis was used to assess each score’s discriminatory performance, quantified by AUC value. AUC ≥ 0.9 was considered excellent, 0.8-0.89 very good, 0.7-0.79 good, 0.6-0.69 moderate, and < 0.6 poor[19]. Model calibration was evaluated using the Hosmer-Lemeshow goodness-of-fit test[20]. Youden index was employed to determine the optimal cutoff for each score, followed by calculation of the sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy. The AUCs of various scores were compared using the DeLong test[20]. We also conducted sensitivity analyses to compare the performance of scores among those < 33 and ≥ 33 gestational weeks. P value (two-tailed) < 0.05 was considered statistically significant.
RESULTS
The study flowchart is shown in Figure 1. Of 625 neonates screened over the 15-month study period, 416 met the inclusion criteria. After excluding 117 for different reasons, the final cohort included 299 neonates. The characteristics of study participants are shown in Table 1. The mean gestational age was 33.3 weeks, with a third (34.8%) being < 33 weeks, and the median birth weight was 1.9 kg. Most participants (80%) were born by cesarean section, and one-quarter (26.1%) required resuscitation at birth. The majority (83.6%) experienced respiratory system affection. Non-invasive respiratory support and invasive ventilation were needed for 70.9% and 35.8% of cases, respectively. The incidence of death during NICU stay was 35.1%. Compared with survivors, non-survivors showed significantly lower mean gestational age (31.5 weeks vs 34.3 weeks), median birth weight (1.5 kg vs 2.1 kg), and median APGAR score at 5 minutes (6 vs 8), and they were more likely to require neonatal resuscitation (56.2% vs 9.8%) and invasive ventilation (92.4% vs 5.2%).
Neonates who passed away exhibited significantly lower MSNS score and higher values on all other scores. The distribution of scores in relation to mortality is summarized in Figure 2. Overall, neonatal mortality escalated in parallel with reduced MSNS score and the increase in the remainder of scores. Each point decrease in MSNS was associated with 2.1 (95% confidence interval: 1.8-2.5) higher odds of mortality. For each point increment in the remainder scores, the odds (95% confidence interval) of mortality increased by 1.6 (1.4-1.8) for CRIB II, 1.2 (1.15-1.26) for SNAP-II, 1.2 (1.14-1.23) for SNAPPE-II, 3.9 (2.8-5.5) for STOPS, and 2.2 (1.8-2.6) for nSOFA.
Figure 2 Percentages of neonatal mortality by scoring system categories.
CRIB II: Clinical Risk Index for Babies II; SNAP-II: Score for Neonatal Acute Physiology II; SNAPPE-II: Score for Neonatal Acute Physiology Perinatal Extension II; STOPS: Sensorium, temperature, oxygenation, perfusion, skin color, and blood sugar; MSNS: Modified Sick Neonatal Score; nSOFA: Neonatal Sequential Organ Failure Assessment.
Table 2 presents the predictive performance of scoring systems for in-hospital neonatal mortality in the overall analysis, and the detailed results for each score are provided in Supplementary Tables 1-6 and Supplementary Figures 1-10. The AUC values were MSNS (0.91) > SNAPPE-II and STOPS (each 0.89) > nSOFA and SNAP-II (each 0.88) > CRIB II (0.82) (Figure 3). The overall accuracy ranged from 87.3% for each of MSNS and STOPS (at cutoffs < 10 and > 2, respectively) to 75.3% for CRIB II (at cutoff > 4). All models showed good calibration (Hosmer-Lemeshow test P > 0.05). The DeLong test showed no significant differences in AUCs of various scores, except for CRIB II which had significantly lower value (Table 3).
Stratified analyses by gestational age (< 33 and ≥ 33 weeks) are provided in Supplementary Tables 7-10. The results were generally consistent with the overall analysis, except CRIB II performed better in neonates < 33 weeks with AUC comparable to that of most other scores.
DISCUSSION
This is the first study to simultaneously compare the predictive accuracy of six neonatal mortality scoring systems (CRIB II, SNAP-II, SNAPPE-II, STOPS, MSNS, and nSOFA) in a single cohort from a high-mortality, resource-limited setting. All six scores demonstrated good to excellent discriminatory ability for in-hospital neonatal mortality, with the MSNS exhibiting the highest overall performance. These findings support the use of more feasible yet reliable neonatal mortality prediction scores in resource-limited settings.
The MSNS, with an AUC of 0.91, emerged as the most accurate score for discriminating neonatal mortality in our study. Its integration of readily accessible clinical signs and perinatal variables without reliance on invasive or costly investigations makes it especially practical for resource-limited settings[9]. This finding aligns with previous Indian studies, reporting AUC values ranging from 0.88 to 0.98 for MSNS[9,21-23]. Similarly, STOPS, which uses only clinical examination and bedside blood glucose without perinatal factors, demonstrated good predictive accuracy (AUC: 0.89) in our cohort. Its rapid applicability at NICU admission makes it particularly useful for critically ill, outborn neonates. This finding is consistent with a recent multicenter Indian study involving neonates ≥ 33 weeks’ gestation, showing an excellent discriminatory ability for STOPS (AUC: 0.92)[11]. Notably, the current study is the first to validate MSNS and STOPS in a non-Indian population and to extend STOPS validation to neonates < 33 weeks’ gestation.
nSOFA, originally designed to assess organ dysfunction and sepsis-related outcomes in preterm babies[10], also showed excellent predictive performance in our general NICU population (AUC: 0.88). This finding is in line with large-scale studies supporting its role in predicting both sepsis-related and all-cause neonatal mortality across gestational ages (AUC: 0.88-0.89)[10,24,25]. The nSOFA score is calculated based on the SpO2:FiO2 ratio and presence of mechanical ventilation for respiratory dysfunction, use of vasoactive agents and systemic corticosteroids for cardiovascular dysfunction, and presence and degree of thrombocytopenia for hematologic dysfunction[10]. Its organ dysfunction-based structure, unlike CRIB II and SNAPPE-II which emphasize perinatal factors, allows clinical application of nSOFA across a range of clinical conditions and gestational ages[10,24-26]. However, dependence on FiO2 and inotrope use introduces subjectivity and potential for institutional bias[7,24].
In the current study, SNAPPE-II and SNAP-II exhibited high discrimination (AUCs: 0.88-0.89) and best sensitivity (83%-87.6%). Their use of physiological parameters, including mean blood pressure, urine output, pO2/FiO2 ratio, and pH, likely enhances sensitivity to detect critical illness[8,13]. While SNAPPE-II and SNAP-II remain robust options in well-equipped NICUs, their reliance on difficult-to-measure clinical variables (e.g., mean blood pressure and urine output) and arterial blood gas analysis may limit applicability in resource-limited units[11,13,15]. The demand for blood gas tests also applies to CRIB II, which performed less favorably overall in our cohort but showed predictive accuracy comparable to other scores among neonates < 33 weeks, reaffirming its utility in preterm populations[8,12,15,16]. The lower performance in neonates ≥ 33 (AUC: 0.72) compared with those < 33 weeks’ gestation (AUC: 0.82) is likely attributed to the limited inclusion of postnatal physiological parameters and heavy weighting of perinatal variables[8].
Several studies have investigated the predictive accuracy of neonatal mortality scores either individually or through pairwise comparison. However, only a limited number of studies have conducted direct head-to-head comparisons involving three or more scoring systems. For instance, in a retrospective study of 759 preterm neonates, Hao et al[17] reported that nSOFA demonstrated the highest predictive accuracy for mortality (AUC: 0.90), outperforming SNAPPE-II (AUC: 0.82) and CRIB-II (AUC: 0.79). In another Chinese study of extremely low birthweight infants, CRIB (AUC: 0.787) showed superior mortality prediction compared with CRIB II (AUC: 0.730), SNAPPE-II (AUC: 0.688), neonatal critical illness severity score (AUC: 0.667), and SNAP-II (0.653) in mortality prediction[27]. Similarly, Gagliardi et al[15] found that CRIB and CRIB-II had better mortality discrimination than SNAPPE-II (AUCs: 0.90 and 0.91 vs 0.84, respectively) among low birthweight newborns. In contrast, a Japanese study of extremely preterm babies reported comparable predictive performance for SNAPPE-II (AUC: 0.95), CRIB II (AUC: 0.93), and SNAP-II (0.90)[28]. Likewise, a Brazilian study demonstrated no significant differences in neonatal mortality prediction performance among CRIB, SNAP, SNAP-II, SNAPPE, and SNAPPE-II[29]. Two recent systematic reviews highlighted the high accuracy of SNAPPE-II in predicting neonatal mortality across general and preterm NICU populations; however, newer tools such as STOPS, MSNS, and nSOFA were not included in these evaluations[5,14].
The selection of a neonatal mortality prediction score must balance discriminatory accuracy and operational feasibility[5,7,8]. Scores combining more sophisticated clinical and laboratory factors are appropriate in high-resource settings with readily accessible diagnostic facilities and electronic health record integration. In contrast, scores like MSNS and STOPS offer practical advantages by requiring minimal training and no invasive investigations, rendering them suitable for NICUs in LMICs with high patient volumes and limited staff and resources[5,11,30]. Our finding that simple, bedside clinical scores such as MSNS and STOPS achieve predictive accuracy comparable to more complex scores like SNAPPE-II is particularly relevant for resource-limited settings, where the use of simple scores is not just a matter of convenience but a necessity due to limited access to laboratory testing[8,14,31].
The high mortality rate observed in our cohort (35.1%) reflects the case mix of our NICU as a major referral center receiving a large proportion of critically ill, outborn neonates, yet operates with limited staff and resources. Although this rate is higher than that reported in most high-income settings[12,13,25], it is consistent with data from some Egyptian and LMICs NICUs[32-37]. These findings reaffirm the validity of neonatal mortality prediction scores even in high-mortality, resource-limited settings. This resource-limited context emphasizes the importance of integrating these scoring tools into the routine medical practice in NICU to aid early clinical decision-making and optimize resource allocation. Their adoption, along with broader health system strengthening, can help reduce neonatal mortality in LMICs[9,11,30]. The high illness severity and poor prognosis also explain the lower cutoff values of mortality prediction scoring systems in our cohort (except for MSNS) compared with prior studies[5,11,17,33]. Indeed, the variability in scoring performance across studies underscores the importance of context-specific score calibration[13,33,38].
A major strength of the present study is its prospective design and simultaneous evaluation of multiple scores in a single population receiving similar management protocols, allowing for direct head-to-head score comparison. Moreover, the study extends the validation of MSNS and STOPS to new populations. Additionally, the findings gain particular importance as they originate from a high-volume NICU in an LMIC, where data collection on neonatal care is usually challenging due to defective documentation of assessments, investigations, and therapies[3,30,39]. Nevertheless, we acknowledge some limitations.
First, the assessment of scores at a single time point does not capture dynamic changes in a patient’s clinical condition over time. Repeated measurements may enhance accuracy, particularly for scores heavily relying on postnatal factors such as STOPS and nSOFA. In a retrospective study of preterm neonates with suspected late-onset sepsis, Kurul et al[40] reported that the nSOFA score measured at 6 hours after initial evaluation had higher discriminative ability for predicting 10-day mortality than the score at initial evaluation (AUC: 0.72 vs 0.69). Similarly, in a retrospective study of newborns ≤ 32 weeks’ gestation with culture-confirmed late onset sepsis, nSOFA scores increased significantly over the first 24 hours among non-survivors (median score 8 at onset; 11 at 6 hours, 11 at 12 hours, and 12 at 24 hours), whereas no such trend was observed among survivors[41]. In an electronic medical record-based study of neonates < 33 weeks’ gestation with late-onset sepsis, nSOFA scores showed a steady rise beginning 12 hours before sepsis evaluation, in non-survivors, with a more pronounced increase among those with early mortality[42]. Likewise, a Brazilian cohort study of VLBW newborns with late-onset sepsis reported a consistent increase in the odds of mortality when nSOFA was measured at 12, 24, and 48 hours after sepsis diagnosis compared with earlier measurements[43]. Finally, Wynn and Polin[10] found improved mortality prediction with serial nSOFA assessments in VLBW newborns with late-onset sepsis (AUC: 0.77 at initial evaluation, 0.78 at 6 hours, and 0.93 at 12 hours). In contrast, Anne et al[44] reported that changes in SNAP-II over 72 hours offer no superior discrimination or sepsis-related mortality in preterm infants compared with the initial score alone. Taken together, scores based on dynamic postnatal variables like nSOFA translate the escalation of clinical care into a quantifiable predicted risk[42]. However, this approach has the potential to introduce treatment-related bias[24,42]. For example, the criteria for initiating vasopressors or corticosteroids to maintain blood pressure in newborns remain controversial and vary across centers[45].
Additional limitations include the single-center design and the high mortality rate, which may limit the generalizability of the findings to high-resource and low-mortality settings. Furthermore, the study focused only on in-hospital mortality outcomes without including long-term morbidities, such as neurodevelopmental impairment and bronchopulmonary dysplasia. Finally, the relatively small sample size rendered the study underpowered for robust subgroup analyses, such as for etiology-specific mortality.
CONCLUSION
The current study demonstrated equivalent good-to-excellent performance of six neonatal mortality prediction scores in a high-mortality, resource-limited NICU setting. Among them, MSNS, STOPS, and nSOFA represent practical and reliable tools that can be integrated into routine clinical care in LMICs. Future comparative, multicenter, longitudinal studies are warranted to further assess the utility of these scores in predicting both short- and long-term neonatal outcomes.
ACKNOWLEDGEMENTS
The authors thank the healthcare professionals who managed sick neonates participating in this study.
Vardhelli V, Murki S, Tandur B, Saha B, Oleti TP, Deshabhotla S, Mohammed YA, Seth S, Siramshetty S, Kallem VR. Comparison of CRIB-II with SNAPPE-II for predicting survival and morbidities before hospital discharge in neonates with gestation ≤ 32 weeks: a prospective multicentric observational study.Eur J Pediatr. 2022;181:2831-2838.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 1][Cited by in RCA: 10][Article Influence: 2.5][Reference Citation Analysis (0)]
DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach.Biometrics. 1988;44:837-845.
[PubMed] [DOI] [Full Text]
Fleiss N, Coggins SA, Lewis AN, Zeigler A, Cooksey KE, Walker LA, Husain AN, de Jong BS, Wallman-Stokes A, Alrifai MW, Visser DH, Good M, Sullivan B, Polin RA, Martin CR, Wynn JL. Evaluation of the Neonatal Sequential Organ Failure Assessment and Mortality Risk in Preterm Infants With Late-Onset Infection.JAMA Netw Open. 2021;4:e2036518.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 88][Cited by in RCA: 86][Article Influence: 17.2][Reference Citation Analysis (1)]
Lewis AN, de la Cruz D, Wynn JL, Frazer LC, Yakah W, Martin CR, Yang H, Itriago E, Unger J, Hair AB, Miele J, Sullivan BA, Husain A, Good M. Evaluation of the Neonatal Sequential Organ Failure Assessment and Mortality Risk in Preterm Infants with Necrotizing Enterocolitis.Neonatology. 2022;119:334-344.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 5][Cited by in RCA: 24][Article Influence: 6.0][Reference Citation Analysis (0)]
Sokou R, Tritzali M, Piovani D, Konstantinidi A, Tsantes AG, Ioakeimidis G, Lampridou M, Parastatidou S, Iacovidou N, Kokoris S, Nikolopoulos GK, Kopterides P, Bonovas S, Tsantes AE. Comparative Performance of Four Established Neonatal Disease Scoring Systems in Predicting In-Hospital Mortality and the Potential Role of Thromboelastometry.Diagnostics (Basel). 2021;11:1955.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 2][Cited by in RCA: 16][Article Influence: 3.2][Reference Citation Analysis (0)]
Khallaf MKT, El-Mahdy HS, Shihab NS, El-Shafiey RMG. Comparison between Score for Neonatal Acute Physiology-Perinatal Extension II (SNAP-PE II) and Clinical Risk Index for Babies (CRIB) in Prediction of Neonatal Mortality in Neonatal Intensive Care Unit (NICU) of Tanta University Hospital.Med J Cairo Univ. 2018;86:3873-3879.
[PubMed] [DOI] [Full Text]
Asker HS, Satar M, Yıldızdaş HY, Mutlu B, Özyurt BM, İpek MŞ, Sivaslı E, Taviloğlu Ş, Çelik Y, Özcan K, Burgut R, Ünal İ. Evaluation of Score for Neonatal Acute Physiology and Perinatal Extension II and Clinical Risk Index for Babies with additional parameters.Pediatr Int. 2016;58:984-987.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 11][Cited by in RCA: 10][Article Influence: 1.0][Reference Citation Analysis (0)]
